This comes up from time to time. Here's another thread where Dennis chimed in
. From what I can tell, a side array (i.e. multiple rows of side surrounds) requires the signals being passed to the subsequent set of surrounds be decorrelated from the first. Those in the know are rather tight lipped about the best method to do this, but Toole's book recommends a time delay on the order of 10 ms. I've also read some papers on commercial cinema calibration where a similar spec is used to calibrate multiple rows of surrounds.
Further, there are some papers discussing the psychoacoustics associated with different methods of decorrelation (see the thread linked to above) that utilized FIR filtering to decorrelate the signals. The theory is by utilizing an all pass filter that randomizes the phase of the second signal with respect to the first the two signals can be decorrelated. The effect is to reduce the perceived comb filtering and increase the perceived size of the room. The latter is an attempt to replicate the complex nature of the phase of multiple signals interacting with objects in a large space.
Unfortunately, FIR filters are computationally expensive, and generally speaking require higher end hardware. Most of the less expensive DSPs only have the ability to implement IIR filters. Dennis generally specs high end DSPs capable of implementing FIR filters, but no word on whether or not they actually use them for this purpose or not.